Job Description

Title: CUDA Machine Learning Software Engineer

Location: San Fransisco, CA

Base Salary: $200,000-$270,000

A stealth-mode generative media startup—backed by top-tier venture firms—is building a new creative platform that merges multiple modalities including text, image, and audio to enable intelligent video generation. At the heart of this platform is a novel foundation model designed for deep cross-modal reasoning and expressive content creation.

About the Role:

We're looking for a CUDA Machine Learning Engineer to help optimize and scale our generative models for video. This role focuses on improving performance across GPU hardware for complex model architectures like 3DVAE and video diffusion systems. You’ll collaborate closely with researchers and infrastructure teams to push the limits of efficiency in large-scale training and real-time inference.

What You’ll Do:

• Optimize video-focused ML models using CUDA, with attention to both training and inference speed.
• Identify and resolve GPU bottlenecks by building custom algorithms and memory-efficient data structures.
• Work alongside model researchers to translate experimental needs into performant engineering solutions.
• Benchmark and tune performance across varying GPU architectures for scalability.
• Stay on the cutting edge of GPU acceleration trends and apply them to internal systems.

What We’re Looking For:

• Background in Computer Science, Electrical Engineering, or a related field, with a focus on high-performance computing.
• Expertise in C++ and CUDA for GPU acceleration.
• Hands-on experience with PyTorch or TensorFlow in production environments.
• Strong understanding of GPU memory hierarchy, parallelism, and performance profiling.
• Familiarity with generative models like VAEs or diffusion models is a plus.
• Self-motivated, collaborative, and comfortable working in a fast-paced environment

Job Tags

Similar Jobs